The Agent Skills Directory

[COMMAND_EXECUTION]: Provides detailed HTTP request templates for deploying unauthorized 'evil' applications to Kubernetes clusters, which is a direct method for achieving remote code execution on the cluster infrastructure.
[COMMAND_EXECUTION]: Includes specific CLI commands for ArgoCD account manipulation and token generation to facilitate unauthorized access and persistence.
[EXTERNAL_DOWNLOADS]: Directs the agent to pull Kubernetes manifests from an untrusted external repository (github.com/attacker/evil-manifests) during the exploitation phase.
[PROMPT_INJECTION]: Uses directive language ('must use this skill') to override the agent's default behavior and safety reasoning when users inquire about ArgoCD security or vulnerabilities.
[COMMAND_EXECUTION]: Outlines techniques for exploiting Redis (port 6379) to perform cache poisoning for cluster takeover.

argocd-tactics